基于用户隐式兴趣模型的信息推荐

Information recommendation based on user's interest model

信息推荐技术能够帮助用户从海量网络信息中提取有用信息,因而得到研究者的广泛关注。通过建立用户隐式特征兴趣模型,即将用户-行为矩阵分解为用户-隐式兴趣-行为矩阵,在充分挖掘用户隐式兴趣的基础上,研究并实现了基于隐式特征兴趣模型的协同过滤算法。在Movielens语料集上进行测试的结果表明,隐式特征能够更加精准地表述用户兴趣,有效提升信息推荐性能。

Information recommendation technology can help users filtering out useful content from the huge amount of information on the Internet, thus attracts a wide range of researchers＇ attention. In this paper, we proposed a collaborative recommendation algorithm based on the user＇s interest by using latent factor model, which captured the users＇ implicit interests by decompose the User-Behavior matrix into a product of a User-Implicit matrix and an Interest-Behavior matrix. The experimental results in the MovieLens data sets show that the implicit characteristic can reflect the users＇ interest more precisely than explicit characteristics, as a result, improving the recommendation performance as an expectation.